Laser ultrasonic imaging of complex defects with full-matrix capture and deep-learning extraction

Area of Science:

• Materials Science
• Non-Destructive Evaluation
• Ultrasonic Imaging

Background:

• Traditional phased array ultrasonic imaging requires couplant, limiting use in hazardous environments and online monitoring.
• Laser ultrasonic (LU) technique offers non-contact inspection but suffers from low sensitivity and complex wave mode conversion.

Purpose of the Study:

• To develop laser-induced full-matrix ultrasonic imaging for complex defect characterization.
• To enhance LU imaging performance using full matrix data acquisition and deep learning.

Main Methods:

• Adapted full matrix data acquisition and deep learning algorithms to the laser ultrasonic technique.
• Conducted simulations and experiments on aluminum samples with representative defects.
• Utilized end-to-end deep learning networks for image reconstruction and quantitative analysis.

Main Results:

• Achieved excellent imaging performance for complex defects using the proposed laser ultrasonic method.
• Demonstrated good agreement between numerical simulations and experimental results.
• The deep learning method produced superior images with quantitative information compared to the traditional total focusing method.

Conclusions:

• The proposed couplant-free laser ultrasonic imaging method effectively visualizes complex defects.
• Deep learning significantly enhances imaging quality and provides quantitative defect assessment.
• This technique shows promise for early-stage defect detection in hazardous environments and in-situ manufacturing monitoring.